kb-health

Pass

Audited by Gen Agent Trust Hub on Apr 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill is instructed to execute a local Python script (../obsidian-notes-karpathy/scripts/lint_obsidian_mechanics.py) to establish a baseline for its health checks, introducing a dependency on external script execution.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and processing data from numerous files within the wiki.
  • Ingestion points: Files located in wiki/live/, wiki/briefings/, outputs/qa/, outputs/content/, outputs/reviews/, and outputs/health/.
  • Boundary markers: Absent; the agent lacks instructions to ignore potential commands embedded in analyzed wiki content.
  • Capability inventory: The agent is authorized to write file reports and perform mechanical fixes in several directories, as well as execute a maintenance script.
  • Sanitization: No content sanitization or validation is specified before processing or modifying data.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 8, 2026, 07:53 AM